Choose email security for Microsoft 365 by measuring what still reaches inboxes after the built-in filter, which carries a spam effectiveness SLA above 99%¹. AegisAI is one example of an add-on, and it reads mail through the Microsoft 365 API after delivery¹⁵.

What you can do by the end

What does Microsoft 365 include, and where does native protection stop?

Every Microsoft 365 tenant includes built-in protection for cloud mailboxes. Microsoft lists a false positive ratio SLA below 1:250,000 for it² and a monthly uptime SLA of 99.999%. Defender for Office 365 Plan 1 adds protection against zero-day malware and phishing, including business email compromise⁴. Microsoft prices Plan 1 at $2.00 user/month, paid yearly⁶. Plan 2 adds investigation and response after a breach, plus phishing simulations.

AegisAI's inbound email security page prints one figure for that mail: 3 in 10 confirmed phishing emails cleared native Gmail and Microsoft filters in Q3 2026²¹. Count your own number in a trial before you buy anything. The page on which Microsoft 365 E3 and E5 email security features you already have maps the licences.

Choose the right deployment model: SEG or API-native

Pick the model by what you will change in mail routing. A secure email gateway (SEG) needs an MX record change, so every message passes through it before Exchange Online delivers it. An API-native add-on connects to the tenant through Microsoft Graph and reads mail that has already arrived. Routing stays as it is.

The model also sets the rollback. Undoing a gateway means restoring the earlier MX records. Undoing an API add-on means revoking its app permissions. The guide on how to replace a secure email gateway without a cutover weekend covers running both models side by side.

Question to ask: does deployment touch our MX records, and what are the exact rollback steps?

Check what the connection can read in your tenant

An API add-on reads what its Graph permissions allow, so read them before you consent. The Mail.Read application permission lets an app read mail in all mailboxes without a signed-in user, and it needs admin consent¹⁰. Removing mail needs a write permission on top of that. Grant it after the monitoring period, as a separate decision.

An admin can revoke permissions granted for the whole organization. In Entra ID, the admin opens the enterprise application, picks the permission on the Admin consent tab and selects Revoke permission¹¹.

Question to ask: which Graph permissions do you request, which of them can move or delete mail, and when do you need each one?

Ask what happens to mail when the add-on is offline

Mail should keep arriving when the add-on is down. The built-in filter carries its own uptime SLA. An API add-on reads mail after Exchange Online has delivered it, so an outage on its side pauses detection and leaves delivery alone. A gateway sits in the MX path, so inbound mail depends on the gateway being up.

Question to ask: if your service is unavailable for an hour, where is our inbound mail, and what do you scan when you come back?

Ask for a reason with every verdict

Ask for a plain-language reason with every block and every removal. The reason should name what the message asked for and the signal that decided the call. Read the reasons for one week of trial verdicts before you turn on removal.

Question to ask: which reason and signals decided the call on one removed message and one allowed message from our trial?

What should the best account takeover protection for Microsoft 365 watch?

It should watch sign-ins and Inbox rule changes together. In the Outlook rules attack, the attacker steals a user's credentials, signs in to the mailbox and creates a forwarding Inbox rule¹². Business email compromise losses reported to IC3 in 2025 were $3,046,598,558¹³.

Identity data has licence limits. Full details on most Entra ID Protection risk detections need Microsoft Entra ID P2; without it, detections arrive as "Additional risk detected" with no detail⁹. The page on how email account takeover works walks through the attack.

Question to ask: which sign-in data do you read, which Entra ID licence does it need, and do you watch Inbox rule changes?

Which compliance and audit readiness checks matter for regulated industries?

The checks that matter are log retention and evidence of each action. In Audit (Standard), Microsoft Purview retains audit records for 180 days⁷. Audit (Premium) keeps Entra ID and Exchange records for one year by default. Ask whether an add-on keeps its own record of each verdict and each removal. Those records need a retention period that fits your rules and an export your auditors can read.

Question to ask: how long do you keep verdict and removal logs, and can we export them to our SIEM?

How to cut management overhead for lean IT teams

Cut overhead by choosing an add-on that needs no routing work and explains its own verdicts. Start in monitoring mode, so the add-on reports and nothing moves. Turn on removal once the reasons hold up. Ask how a user report reaches the tool, and whether one removal reaches every copy of a message.

Question to ask: how many hours a week does an admin spend in your console after the first month, and on what tasks?

Which key evaluation criteria for M365 email security solutions does each approach meet?

How to test the best email security for Microsoft 365 in a vendor-neutral assessment

Answer these 10 questions yes or no with evidence from a trial on your own tenant.

How AegisAI handles this

Reads mail through the Microsoft 365 or Google Workspace API after delivery¹⁵. Runs in monitoring mode first. Removal needs write access your admin grants and can revoke. The same API connection as inbound, plus audit-log access²⁴. On Microsoft 365, sign-in reports need Entra ID P1.

How to measure it in 30 days

Track three numbers from the first day of the trial.

Questions people ask

Sources